Jiang Chenshan, Cheng Wenjie, Jiang Xinyi, Zhang Jianlin, Tang Xiaojun
Department of Maxillo-facial Surgery, Plastic Surgery Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China.
Department of Ear Reconstruction, Plastic Surgery Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China.
J Craniofac Surg. 2025 Jul 31. doi: 10.1097/SCS.0000000000011698.
Mandibular angle osteotomy (MAO) is one of the most effective ways to correct square facial contours. With the development of Artificial Intelligence (AI) technology, particularly in medicine, more patients are seeking medical queries from online websites. This study compared the performance of 2 AI platforms, ChatGPT-4o and DeepSeek in answering questions about MAO.
Twenty frequently asked questions about MAO were selected and answered by ChatGPT-4o and DeepSeek. The responses from 2 platforms were graded by 9 experienced craniomaxillofacial plastic surgeons from 2 different hospitals. The relevance, accuracy, completeness, and readability of responses were evaluated. The 20 questions were divided into 4 categories: general conception, surgery process, complication, and other topics. Statistical analysis, including the 2-sided t test and Kruskal-Wallis test was applied to compare metrics.
Both ChatGPT-4o and DeepSeek provided high-quality information about MAO. However, ChatGPT-4o outperformed in giving more thorough answers (4.4945±0.03089 vs. 4.4315±0.02519, P=0.048), and DeepSeek outperformed in giving answers more easily to read (4.2960±0.04717 vs. 4.1965±0.03986, P=0.026). Also, although ChatGPT performed well in answering all kinds of questions, DeepSeek had weak performance in answering questions regarding surgery process of MAO.
Both platforms offered reliable information. Compared to DeepSeek, ChatGPT-4o provided more thorough responses and was more aligned with clinical practice. This study discovered the potential of AI platforms in addressing patient education and providing medical information in craniomaxillofacial plastic surgery field.
下颌角截骨术（MAO）是矫正方形面部轮廓最有效的方法之一。随着人工智能（AI）技术的发展，尤其是在医学领域，越来越多的患者在在线网站上寻求医疗咨询。本研究比较了两个AI平台ChatGPT-4o和豆包在回答有关MAO问题方面的表现。
选择了20个关于MAO的常见问题，由ChatGPT-4o和豆包进行回答。来自两家不同医院的9位经验丰富的颅颌面整形外科医生对两个平台的回答进行评分。评估回答的相关性、准确性、完整性和可读性。这20个问题分为4类：一般概念、手术过程、并发症和其他主题。应用双侧t检验和Kruskal-Wallis检验等统计分析方法比较各项指标。
ChatGPT-4o和豆包都提供了关于MAO的高质量信息。然而，ChatGPT-4o在给出更全面的答案方面表现更优（4.4945±0.03089对4.4315±0.02519，P=0.048），而豆包在给出更易读的答案方面表现更优（4.2960±0.04717对4.1965±0.03986，P=0.026）。此外，尽管ChatGPT在回答各类问题方面表现良好，但豆包在回答关于MAO手术过程的问题时表现较弱。
两个平台都提供了可靠的信息。与豆包相比，ChatGPT-4o提供了更全面的回答，并且与临床实践更相符。本研究发现了AI平台在颅颌面整形外科领域进行患者教育和提供医疗信息方面的潜力。
